Over the recent years, with an increase in data quantity treated by an enterprise, a large scale of an online system and an introduction of cloud computing have been accelerated. Key-Value Store (KVS) is one of techniques for processing data fast. The KVS is a technique for saving and managing the data. The KVS is configured to manage the data as a pair of data (Value) desired to be saved and a unique identifier (Key) associated therewith. The KVS acquires the data with a mechanism as simple as fetching Value by designating Key in response to a processing request given from a client in the online system and other equivalent systems. The data can be saved in distribution to a plurality of servers corresponding to Keys, and therefore the servers are easy to be extended. In other words, the KVS exhibits excellent extensibility and reduces a delay of processing time. Speed-up of the processing is facilitated by extending the servers. Note that a scheme for expanding a system scale by extending the servers is said to be “Scale Out”.
A distributed KVS (DKVS) configured to save the data by distributing the data to a plurality of servers is suited to a stateless processing method, and is easy to attain the scale-out. The stateless processing method is a method of determining output data with respect to input data without retaining data indicating a present state and other equivalent states.
On the other hand, a task service and other equivalent services by the online system are generally hard to be attained by only the stateless processing method, and particularly an important task service adopts a stateful processing method. The stateful processing method is a method of retaining the data indicating the present state and other equivalent data, and determining the output data by reflecting a content of the retained data in the input data. When adding functions instanced by a transaction, exclusive control, assurance of reading consistency and a countermeasure against split-brain syndrome in a Relational Database Management System (RDBMS) and other equivalent systems to the stateless extensible distributed KVS, there is no alternative but to manage the states, and the processing method results in being stateful with no extensibility.
When increasingly flexible to the transaction, the exclusive control, the assurance of reading consistency and the countermeasure against split-brain syndrome with respect to the task service of the distributed KVS and other equivalent services, a stateless property is lost, and the processing method results in being stateful with no extensibility. Accordingly, the service does not become eligible as the task service of the distributed KVS, and hence the extensibility is lost.